Abstract Single-cell RNA-sequencing (scRNA-seq) has emerged as a revolutionary tool that allows us to address scientific questions that were elusive just a few years ago. Specifically, the scRNA-seq technology has the potential to enable insights into how a single cell develops into a complex organism, how cell-to-cell variation functions to maintain homeostasis, and how diseases such as cancer develop resistance to treatment. Unfortunately, much of this potential has yet to be realized as statistical methods to analyze scRNA-seq data are lacking. For many types of analyses, the methods currently in use obscure and, in some cases, distort biological signals. A number of statistical and computational challenges must be addressed to prevent inaccurate conclusions, and to optimize novel discovery. This proposal addresses those challenges. In particular, we propose normalization methods to adjust scRNA-seq data for artifacts and thereby ensure robust and accurate downstream inference. We also propose statistical methods to characterize and remove variability induced by oscillatory genes which can mask effects under study. By de-noising the data, these methods will improve the power with which signals of interest can be studied. Finally, we propose methods for identifying cell sub-populations and characterizing their differences across biological conditions. These types of methods are required for identifying novel sub-populations and for characterizing the ways in which cell- specific expression changes in response to various environments. Successful completion of the project will help to ensure that maximal information is obtained from powerful scRNA-seq experiments.